Googling hidden interactions: Network construction from WWW

 

HAWOONG JEONG

COMPLEX SYSTEMS AND STATISTICAL PHYSICS LAB

 

Massive digital records have made it possible to analyze an enormous amount of data in various research fields, such as social network analysis and systems biology. We investigate weighted relatedness networks by extracting information on the World Wide Web. Using famous search engines such as Google, we quantify the relatedness between two objects as the number of webpages including both of their names and construct weighted relatedness networks. We take some representative examples in relatedness networks among people, measure the distributions of quantities of interest in weighted network analysis,and present a class of measure called Renyi disparity, which characterizes the homogeneity of weight distribution for an individual node. The concept of maximum relatedness subnetwork, which captures the most essential relation for each individual, is also introduced. There is a tremendous amount of data on the Web, which can prove very useful ones if we harness it cleverly. Search engines are a basic device to classify such information and we have constructed social networks based on the Google correlation values quantifying the relatedness of people. We have verified that our method can discover useful infomation such as hidden correlatioin between any two objects on the Web.